I figured out why the https://t.co/1o6zdY4yzy script was erroring out after the final MAX_TOOL_LOOPS of 9 (9 rounds) and it was because their was no Final Inference Step that saved to the MariaDB database and returned back to the main menu of the https://t.co/X9bY0CEb50 script. I added the required inference step to https://t.co/1o6zdY4yzy and forked the repository to commit a change.
In addition, I found that if the LLM's responses/findings to write to the MariaDB are too long it will also error out. You will want to go into MariaDB and edit the metatron database to increase the column size to allow for longer entries with these commands:
USE metatron_db; -- ← change to your database name
-- Make the result column much larger (safe for descriptive text)
ALTER TABLE exploits_attempted
MODIFY COLUMN result TEXT;
-- Also make notes larger (good practice)
ALTER TABLE exploits_attempted
MODIFY COLUMN notes TEXT;
-- Optional: make payload larger too
ALTER TABLE exploits_attempted
MODIFY COLUMN payload TEXT;
exit;
Hope this helps if anyone else was experiencing these errors.
Just installed this and re-configurate the https://t.co/1o6zdY4yzy script for my Parrot VM to use my host machines Ollama API that uses my GPU instead of installing Ollama inside my VM and allocating it more resources to run the LLM off of CPU.
Its pretty good, I just ran a couple of basic Nmap scans of my private systems. The only thing I noticed so far is that METATRON runs multiple rounds of the LLM query testing for vulnerabilities and exploits and the script seems to always fail after the 9th Round. Is there a way to stop the script/tool calling to the LLM without just ctrl>c out of the Metatron script?
I figured out why the https://t.co/1o6zdY4yzy script was erroring out after the final MAX_TOOL_LOOPS of 9 (9 rounds) and it was because their was no Final Inference Step that saved to the MariaDB database and returned back to the main menu of the https://t.co/X9bY0CEb50 script. I added the required inference step to https://t.co/1o6zdY4yzy and forked the repository to commit a change.
In addition, I found that if the LLM's responses/findings to write to the MariaDB are too long it will also error out. You will want to go into MariaDB and edit the metatron database to increase the column size to allow for longer entries with these commands:
USE metatron_db; -- ← change to your database name
-- Make the result column much larger (safe for descriptive text)
ALTER TABLE exploits_attempted
MODIFY COLUMN result TEXT;
-- Also make notes larger (good practice)
ALTER TABLE exploits_attempted
MODIFY COLUMN notes TEXT;
-- Optional: make payload larger too
ALTER TABLE exploits_attempted
MODIFY COLUMN payload TEXT;
exit;
Hope this helps if anyone else was experiencing these errors. I figured out why the https://t.co/1o6zdY4yzy script was erroring out after the final MAX_TOOL_LOOPS of 9 (9 rounds) and it was because their was no Final Inference Step that saved to the MariaDB database and returned back to the main menu of the https://t.co/X9bY0CEb50 script. I added the required inference step to https://t.co/1o6zdY4yzy and forked the repository to commit a change.
In addition, I found that if the LLM's responses/findings to write to the MariaDB are too long it will also error out. You will want to go into MariaDB and edit the metatron database to increase the column size to allow for longer entries with these commands:
USE metatron_db; -- ← change to your database name
-- Make the result column much larger (safe for descriptive text)
ALTER TABLE exploits_attempted
MODIFY COLUMN result TEXT;
-- Also make notes larger (good practice)
ALTER TABLE exploits_attempted
MODIFY COLUMN notes TEXT;
-- Optional: make payload larger too
ALTER TABLE exploits_attempted
MODIFY COLUMN payload TEXT;
exit;
Hope this helps if anyone else was experiencing these errors.
Just installed this and re-configurate the https://t.co/1o6zdY4yzy script for my Parrot VM to use my host machines Ollama API that uses my GPU instead of installing Ollama inside my VM and allocating it more resources to run the LLM off of CPU.
Its pretty good, I just ran a couple of basic Nmap scans of my private systems. The only thing I noticed so far is that METATRON runs multiple rounds of the LLM query testing for vulnerabilities and exploits and the script seems to always fail after the 9th Round. Is there a way to stop the script/tool calling to the LLM without just ctrl>c out of the Metatron script?
@HuggingModels This model is not good. It produces a ton of command syntax errors and only got a simple request somewhat (not fully) right 1 out of 10 times.